Wastewater treatment plants (WWTPs) are essential for public health but are highly energy-intensive, accounting for approximately 1.6% of global greenhouse gas (GHG) emissions. Developing cost-effective carbon mitigation strategies requires identifying the dominant drivers of these emissions. However, existing analyses often treat operational variables as independent predictors or lack the causal validity required to disentangle complex, interacting emission sources across diverse regions. Consequently, the fundamental mechanisms causing distinct GHG emissions in the resource-constrained wastewater sector remain poorly understood. Here we use an integrated machine learning and causal inference framework to quantify the drivers of GHG emissions across 40,722 Chinese WWTPs over a decade. We reveal that aggregate emission intensities have rebounded recently, driven primarily by increasingly stringent water discharge standards that necessitate chemical- and energy-intensive nitrogen removal. Although ammonia nitrogen removal exerts the strongest individual causal effect on emissions—where a 10% increase drives a 5.65% surge in total intensity—decarbonizing the regional power grid offers a highly potent and practical mitigation lever. Specifically, a 10% reduction in the regional grid electricity emission factor accounts for a 4.45% equivalent reduction in total emissions, demonstrating a profound cross-sectoral coupling. Our findings highlight a critical tension between local water quality improvements and global climate goals, suggesting that future environmental regulations must harmonize ecological benefits with their intrinsic carbon costs.

Sustainable urban wastewater management requires systems that effectively remove pollutants while enabling resource recovery and diminishing environmental impacts. In high-density cities, conventional centralized treatment plants often entail long conveyance distances, high pumping energy, and substantial methane emissions from widespread communal septic tanks, compounded by stringent effluent standards. Scaled decentralized systems (SDSs)—distributed facilities integrated into existing sewer networks—offer potential advantages through reduced transport and localized recovery, yet their performance across scales, technologies, and recovery strategies remains poorly quantified. Here we show that SDSs featuring 20,000 m³ d⁻¹ moving-bed biofilm reactor plants with combined water reuse and additional heat-pump energy recovery achieve the lowest life-cycle environmental impacts and costs. We applied whole-system life-cycle assessment and cost analysis of 29 scenarios treating 100,000 m³ d⁻¹ of wastewater in a city in China. These configurations reduce global warming potential by up to 52.5% relative to an optimized centralized benchmark, owing to shorter sewers that preserve influent carbon for biological denitrification, elimination of external carbon dosing, and efficient dual recovery; upstream septic tanks, however, contribute 24–47% of total warming potential, highlighting a key trade-off. These findings demonstrate that carefully scaled decentralization, paired with robust biofilm technology and integrated recovery, provides a superior pathway for sustainable wastewater infrastructure in dense urban settings.